🚀 Projeto Fase 6 - Entrega 1: Sistema de Visão Computacional¶
📋 Objetivo¶
Desenvolver um sistema de visão computacional usando YOLOv5 customizado para detectar e classificar bicicletas e carros em imagens.
🎯 Metas da Entrega¶
- Dataset organizado: 80 imagens (40 bicicletas + 40 carros)
- Divisão: 32 treino, 4 validação, 4 teste (por classe)
- Rotulação feita no Make Sense IA
- Treinamento com 15 e 40 épocas
- Análise comparativa dos resultados
- Demonstração prática com imagens de teste
📦 1. Instalação e Configuração do Ambiente¶
In [1]:
# Clonar repositório YOLOv5
!git clone https://github.com/ultralytics/yolov5
%cd yolov5
!pip install -r requirements.txt -q
Cloning into 'yolov5'... remote: Enumerating objects: 17611, done. remote: Counting objects: 100% (16/16), done. remote: Compressing objects: 100% (14/14), done. remote: Total 17611 (delta 8), reused 2 (delta 2), pack-reused 17595 (from 3) Receiving objects: 100% (17611/17611), 16.84 MiB | 11.63 MiB/s, done. Resolving deltas: 100% (11984/11984), done. /content/yolov5 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 1.1/1.1 MB 31.5 MB/s eta 0:00:00
In [2]:
# Importar bibliotecas necessárias
import torch
import os
from IPython.display import Image, display
import shutil
from datetime import datetime
In [3]:
# Verificar GPU disponível
print(f'Torch version: {torch.__version__}')
print(f'CUDA disponível: {torch.cuda.is_available()}')
print(f'GPU: {torch.cuda.get_device_name(0) if torch.cuda.is_available() else "CPU"}')
Torch version: 2.8.0+cu126 CUDA disponível: False GPU: CPU
💾 2. Conexão com Google Drive e Organização do Dataset¶
O dataset está organizado da seguinte forma:
/dataset
/train
/images (64 imagens: 32 bicicletas + 32 carros)
/labels (64 arquivos .txt)
/valid
/images (8 imagens: 4 bicicletas + 4 carros)
/labels (8 arquivos .txt)
/test
/images (8 imagens: 4 bicicletas + 4 carros)
In [4]:
# Montar Google Drive
from google.colab import drive
drive.mount('/content/drive')
Mounted at /content/drive
In [5]:
# Verificar estrutura do dataset
base_path = '/content/drive/MyDrive/FIAP/FASE_6/dataset'
# Verificar se o caminho existe
if os.path.exists(base_path):
print("✅ Dataset encontrado!")
print(f"\n📊 Estrutura do dataset:")
# Contar imagens em cada pasta
train_images = len(os.listdir(f'{base_path}/train/images'))
train_labels = len(os.listdir(f'{base_path}/train/labels'))
valid_images = len(os.listdir(f'{base_path}/valid/images'))
valid_labels = len(os.listdir(f'{base_path}/valid/labels'))
test_images = len(os.listdir(f'{base_path}/test/images'))
print(f" Train: {train_images} imagens, {train_labels} labels")
print(f" Valid: {valid_images} imagens, {valid_labels} labels")
print(f" Test: {test_images} imagens")
else:
print("❌ Dataset não encontrado!")
print(f"Verifique o caminho: {base_path}")
✅ Dataset encontrado! 📊 Estrutura do dataset: Train: 64 imagens, 64 labels Valid: 8 imagens, 8 labels Test: 8 imagens
⚙️ 3. Criação do Arquivo de Configuração (dataset.yaml)¶
Este arquivo define:
- Caminhos das pastas de treino e validação
- Número de classes (nc = 2)
- Nomes das classes (bicicleta e carro)
In [6]:
# Criar arquivo dataset.yaml
dataset_yaml_content = f"""
# Dataset configuration for YOLOv5
path: {base_path} # Caminho raiz do dataset
train: train/images # Caminho relativo das imagens de treino
val: valid/images # Caminho relativo das imagens de validação
# Classes
nc: 2 # Número de classes
names: ['bicicleta', 'carro'] # Nomes das classes
"""
# Salvar arquivo yaml
yaml_path = '/content/yolov5/dataset.yaml'
with open(yaml_path, 'w') as f:
f.write(dataset_yaml_content)
print("✅ Arquivo dataset.yaml criado com sucesso!")
print(f"\n📄 Conteúdo:")
print(dataset_yaml_content)
✅ Arquivo dataset.yaml criado com sucesso! 📄 Conteúdo: # Dataset configuration for YOLOv5 path: /content/drive/MyDrive/FIAP/FASE_6/dataset # Caminho raiz do dataset train: train/images # Caminho relativo das imagens de treino val: valid/images # Caminho relativo das imagens de validação # Classes nc: 2 # Número de classes names: ['bicicleta', 'carro'] # Nomes das classes
🏋️ 4. TREINAMENTO 1: 30 ÉPOCAS¶
Neste primeiro treinamento, vamos treinar o modelo YOLOv5 com 15 épocas.
Parâmetros utilizados:¶
- img: 640 (tamanho da imagem)
- batch: 16 (tamanho do lote)
- epochs: 15
- data: dataset.yaml (configuração do dataset)
- weights: yolov5s.pt (pesos pré-treinados do YOLOv5 small)
- cache: Para acelerar o treinamento
In [7]:
# Treinar modelo com 30 épocas
print("🚀 Iniciando treinamento com 15 épocas...")
print(f"⏰ Início: {datetime.now().strftime('%H:%M:%S')}")
%cd /content/yolov5
!python train.py \
--img 640 \
--batch 16 \
--epochs 15 \
--data dataset.yaml \
--weights yolov5s.pt \
--cache \
--name treino_15_epocas
print(f"✅ Treinamento finalizado: {datetime.now().strftime('%H:%M:%S')}")
🚀 Iniciando treinamento com 15 épocas... ⏰ Início: 16:16:22 /content/yolov5 Creating new Ultralytics Settings v0.0.6 file ✅ View Ultralytics Settings with 'yolo settings' or at '/root/.config/Ultralytics/settings.json' Update Settings with 'yolo settings key=value', i.e. 'yolo settings runs_dir=path/to/dir'. For help see https://docs.ultralytics.com/quickstart/#ultralytics-settings. wandb: WARNING ⚠️ wandb is deprecated and will be removed in a future release. See supported integrations at https://github.com/ultralytics/yolov5#integrations. 2025-10-20 16:16:41.163520: E external/local_xla/xla/stream_executor/cuda/cuda_fft.cc:467] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered WARNING: All log messages before absl::InitializeLog() is called are written to STDERR E0000 00:00:1760977001.190015 8594 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered E0000 00:00:1760977001.198357 8594 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered W0000 00:00:1760977001.218863 8594 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1760977001.218907 8594 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1760977001.218915 8594 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1760977001.218919 8594 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. wandb: (1) Create a W&B account wandb: (2) Use an existing W&B account wandb: (3) Don't visualize my results wandb: Enter your choice: (30 second timeout) 3 wandb: You chose "Don't visualize my results" train: weights=yolov5s.pt, cfg=, data=dataset.yaml, hyp=data/hyps/hyp.scratch-low.yaml, epochs=15, batch_size=16, imgsz=640, rect=False, resume=False, nosave=False, noval=False, noautoanchor=False, noplots=False, evolve=None, evolve_population=data/hyps, resume_evolve=None, bucket=, cache=ram, image_weights=False, device=, multi_scale=False, single_cls=False, optimizer=SGD, sync_bn=False, workers=8, project=runs/train, name=treino_15_epocas, exist_ok=False, quad=False, cos_lr=False, label_smoothing=0.0, patience=100, freeze=[0], save_period=-1, seed=0, local_rank=-1, entity=None, upload_dataset=False, bbox_interval=-1, artifact_alias=latest, ndjson_console=False, ndjson_file=False github: up to date with https://github.com/ultralytics/yolov5 ✅ YOLOv5 🚀 v7.0-441-g15c0127a Python-3.12.12 torch-2.8.0+cu126 CPU hyperparameters: lr0=0.01, lrf=0.01, momentum=0.937, weight_decay=0.0005, warmup_epochs=3.0, warmup_momentum=0.8, warmup_bias_lr=0.1, box=0.05, cls=0.5, cls_pw=1.0, obj=1.0, obj_pw=1.0, iou_t=0.2, anchor_t=4.0, fl_gamma=0.0, hsv_h=0.015, hsv_s=0.7, hsv_v=0.4, degrees=0.0, translate=0.1, scale=0.5, shear=0.0, perspective=0.0, flipud=0.0, fliplr=0.5, mosaic=1.0, mixup=0.0, copy_paste=0.0 Comet: run 'pip install comet_ml' to automatically track and visualize YOLOv5 🚀 runs in Comet TensorBoard: Start with 'tensorboard --logdir runs/train', view at http://localhost:6006/ Downloading https://github.com/ultralytics/assets/releases/download/v0.0.0/Arial.ttf to /root/.config/Ultralytics/Arial.ttf... 100% 755k/755k [00:00<00:00, 23.2MB/s] Downloading https://github.com/ultralytics/yolov5/releases/download/v7.0/yolov5s.pt to yolov5s.pt... 100% 14.1M/14.1M [00:00<00:00, 222MB/s] Overriding model.yaml nc=80 with nc=2 from n params module arguments 0 -1 1 3520 models.common.Conv [3, 32, 6, 2, 2] 1 -1 1 18560 models.common.Conv [32, 64, 3, 2] 2 -1 1 18816 models.common.C3 [64, 64, 1] 3 -1 1 73984 models.common.Conv [64, 128, 3, 2] 4 -1 2 115712 models.common.C3 [128, 128, 2] 5 -1 1 295424 models.common.Conv [128, 256, 3, 2] 6 -1 3 625152 models.common.C3 [256, 256, 3] 7 -1 1 1180672 models.common.Conv [256, 512, 3, 2] 8 -1 1 1182720 models.common.C3 [512, 512, 1] 9 -1 1 656896 models.common.SPPF [512, 512, 5] 10 -1 1 131584 models.common.Conv [512, 256, 1, 1] 11 -1 1 0 torch.nn.modules.upsampling.Upsample [None, 2, 'nearest'] 12 [-1, 6] 1 0 models.common.Concat [1] 13 -1 1 361984 models.common.C3 [512, 256, 1, False] 14 -1 1 33024 models.common.Conv [256, 128, 1, 1] 15 -1 1 0 torch.nn.modules.upsampling.Upsample [None, 2, 'nearest'] 16 [-1, 4] 1 0 models.common.Concat [1] 17 -1 1 90880 models.common.C3 [256, 128, 1, False] 18 -1 1 147712 models.common.Conv [128, 128, 3, 2] 19 [-1, 14] 1 0 models.common.Concat [1] 20 -1 1 296448 models.common.C3 [256, 256, 1, False] 21 -1 1 590336 models.common.Conv [256, 256, 3, 2] 22 [-1, 10] 1 0 models.common.Concat [1] 23 -1 1 1182720 models.common.C3 [512, 512, 1, False] 24 [17, 20, 23] 1 18879 models.yolo.Detect [2, [[10, 13, 16, 30, 33, 23], [30, 61, 62, 45, 59, 119], [116, 90, 156, 198, 373, 326]], [128, 256, 512]] Model summary: 214 layers, 7025023 parameters, 7025023 gradients, 16.0 GFLOPs Transferred 343/349 items from yolov5s.pt optimizer: SGD(lr=0.01) with parameter groups 57 weight(decay=0.0), 60 weight(decay=0.0005), 60 bias albumentations: 1 validation error for InitSchema size Field required [type=missing, input_value={'scale': (0.8, 1.0), 'ra...: None, 'strict': False}, input_type=dict] For further information visit https://errors.pydantic.dev/2.11/v/missing train: Scanning /content/drive/MyDrive/FIAP/FASE_6/dataset/train/labels.cache... 64 images, 0 backgrounds, 0 corrupt: 100% 64/64 [00:00<?, ?it/s] train: Caching images (0.1GB ram): 100% 64/64 [01:02<00:00, 1.02it/s] val: Scanning /content/drive/MyDrive/FIAP/FASE_6/dataset/valid/labels.cache... 8 images, 0 backgrounds, 0 corrupt: 100% 8/8 [00:00<?, ?it/s] val: Caching images (0.0GB ram): 100% 8/8 [00:03<00:00, 2.33it/s] AutoAnchor: 3.81 anchors/target, 1.000 Best Possible Recall (BPR). Current anchors are a good fit to dataset ✅ Plotting labels to runs/train/treino_15_epocas/labels.jpg... /content/yolov5/train.py:357: FutureWarning: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. scaler = torch.cuda.amp.GradScaler(enabled=amp) Image sizes 640 train, 640 val Using 2 dataloader workers Logging results to runs/train/treino_15_epocas Starting training for 15 epochs... Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 0/14 0G 0.1256 0.03123 0.03523 34 640: 25% 1/4 [00:48<02:24, 48.02s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 0/14 0G 0.1267 0.03051 0.03385 30 640: 50% 2/4 [01:19<01:16, 38.34s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 0/14 0G 0.1272 0.03027 0.03243 29 640: 75% 3/4 [01:49<00:34, 34.51s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 0/14 0G 0.127 0.02986 0.03252 26 640: 100% 4/4 [02:17<00:00, 34.48s/it] Class Images Instances P R mAP50 mAP50-95: 0% 0/1 [00:00<?, ?it/s]WARNING ⚠️ NMS time limit 0.900s exceeded Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:04<00:00, 4.35s/it] all 8 8 0.00155 0.125 0.00114 0.000447 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 1/14 0G 0.1157 0.02783 0.02899 24 640: 25% 1/4 [00:28<01:25, 28.54s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 1/14 0G 0.1184 0.02846 0.03038 28 640: 50% 2/4 [00:56<00:56, 28.28s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 1/14 0G 0.1181 0.029 0.03129 31 640: 75% 3/4 [01:23<00:27, 27.61s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 1/14 0G 0.1166 0.02924 0.03062 30 640: 100% 4/4 [01:50<00:00, 27.65s/it] Class Images Instances P R mAP50 mAP50-95: 0% 0/1 [00:00<?, ?it/s]WARNING ⚠️ NMS time limit 0.900s exceeded Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:05<00:00, 5.55s/it] all 8 8 0.00313 0.25 0.00291 0.000874 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 2/14 0G 0.1098 0.03044 0.0249 32 640: 25% 1/4 [00:26<01:19, 26.54s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 2/14 0G 0.1046 0.03051 0.02418 31 640: 50% 2/4 [00:53<00:53, 26.69s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 2/14 0G 0.1018 0.03188 0.02377 39 640: 75% 3/4 [01:20<00:26, 26.92s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 2/14 0G 0.09881 0.03152 0.02394 31 640: 100% 4/4 [01:47<00:00, 26.97s/it] Class Images Instances P R mAP50 mAP50-95: 0% 0/1 [00:00<?, ?it/s]WARNING ⚠️ NMS time limit 0.900s exceeded Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:04<00:00, 4.13s/it] all 8 8 0.00352 0.375 0.00941 0.00344 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 3/14 0G 0.09139 0.0329 0.02273 37 640: 25% 1/4 [00:26<01:20, 26.77s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 3/14 0G 0.08161 0.0307 0.02003 24 640: 50% 2/4 [00:53<00:53, 26.90s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 3/14 0G 0.0791 0.02977 0.02218 27 640: 75% 3/4 [01:20<00:26, 26.63s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 3/14 0G 0.08213 0.03045 0.0222 36 640: 100% 4/4 [01:46<00:00, 26.67s/it] Class Images Instances P R mAP50 mAP50-95: 0% 0/1 [00:00<?, ?it/s]WARNING ⚠️ NMS time limit 0.900s exceeded Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:04<00:00, 4.90s/it] all 8 8 0.00348 0.25 0.0207 0.0104 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 4/14 0G 0.09387 0.03083 0.02479 34 640: 25% 1/4 [00:26<01:19, 26.42s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 4/14 0G 0.08957 0.03002 0.02191 30 640: 50% 2/4 [00:53<00:53, 26.82s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 4/14 0G 0.08423 0.03067 0.02037 38 640: 75% 3/4 [01:21<00:27, 27.47s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 4/14 0G 0.08239 0.02977 0.02055 29 640: 100% 4/4 [01:48<00:00, 27.11s/it] Class Images Instances P R mAP50 mAP50-95: 0% 0/1 [00:00<?, ?it/s]WARNING ⚠️ NMS time limit 0.900s exceeded Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:03<00:00, 3.71s/it] all 8 8 0.00349 0.375 0.0777 0.0387 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 5/14 0G 0.06804 0.03374 0.0181 37 640: 25% 1/4 [00:27<01:22, 27.55s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 5/14 0G 0.07521 0.03144 0.01882 33 640: 50% 2/4 [00:57<00:58, 29.05s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 5/14 0G 0.0767 0.03156 0.01956 41 640: 75% 3/4 [01:26<00:28, 28.91s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 5/14 0G 0.07657 0.03208 0.01932 38 640: 100% 4/4 [01:53<00:00, 28.27s/it] Class Images Instances P R mAP50 mAP50-95: 0% 0/1 [00:00<?, ?it/s]WARNING ⚠️ NMS time limit 0.900s exceeded Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:03<00:00, 3.92s/it] all 8 8 0.00334 0.5 0.294 0.151 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 6/14 0G 0.07853 0.03218 0.01842 39 640: 25% 1/4 [00:27<01:21, 27.14s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 6/14 0G 0.07995 0.03169 0.01704 40 640: 50% 2/4 [00:53<00:53, 26.90s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 6/14 0G 0.07989 0.02987 0.0171 30 640: 75% 3/4 [01:21<00:27, 27.25s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 6/14 0G 0.07666 0.02919 0.01815 30 640: 100% 4/4 [01:47<00:00, 26.93s/it] Class Images Instances P R mAP50 mAP50-95: 0% 0/1 [00:00<?, ?it/s]WARNING ⚠️ NMS time limit 0.900s exceeded Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:03<00:00, 3.80s/it] all 8 8 0.00331 0.375 0.137 0.0734 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 7/14 0G 0.06857 0.03056 0.01508 36 640: 25% 1/4 [00:26<01:19, 26.54s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 7/14 0G 0.0727 0.02972 0.01793 39 640: 50% 2/4 [00:53<00:54, 27.01s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 7/14 0G 0.07317 0.02906 0.01778 32 640: 75% 3/4 [01:22<00:27, 27.74s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 7/14 0G 0.07315 0.02907 0.01802 38 640: 100% 4/4 [01:49<00:00, 27.32s/it] Class Images Instances P R mAP50 mAP50-95: 0% 0/1 [00:00<?, ?it/s]WARNING ⚠️ NMS time limit 0.900s exceeded Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:03<00:00, 3.88s/it] all 8 8 0.0658 0.625 0.421 0.203 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 8/14 0G 0.07486 0.03096 0.01938 39 640: 25% 1/4 [00:27<01:21, 27.17s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 8/14 0G 0.07616 0.0271 0.01692 30 640: 50% 2/4 [00:54<00:54, 27.15s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 8/14 0G 0.07259 0.02802 0.01684 38 640: 75% 3/4 [01:22<00:27, 27.41s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 8/14 0G 0.07182 0.02732 0.0161 31 640: 100% 4/4 [01:48<00:00, 27.15s/it] Class Images Instances P R mAP50 mAP50-95: 0% 0/1 [00:00<?, ?it/s]WARNING ⚠️ NMS time limit 0.900s exceeded Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:04<00:00, 4.51s/it] all 8 8 0.0105 0.5 0.269 0.141 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 9/14 0G 0.06313 0.02721 0.0169 35 640: 25% 1/4 [00:26<01:20, 26.92s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 9/14 0G 0.06388 0.02994 0.01434 44 640: 50% 2/4 [00:54<00:54, 27.25s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 9/14 0G 0.06253 0.02914 0.01535 35 640: 75% 3/4 [01:23<00:28, 28.20s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 9/14 0G 0.06236 0.02733 0.01476 27 640: 100% 4/4 [01:51<00:00, 27.90s/it] Class Images Instances P R mAP50 mAP50-95: 0% 0/1 [00:00<?, ?it/s]WARNING ⚠️ NMS time limit 0.900s exceeded Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:03<00:00, 3.85s/it] all 8 8 0.767 0.484 0.652 0.392 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 10/14 0G 0.06493 0.03065 0.01436 39 640: 25% 1/4 [00:27<01:23, 27.77s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 10/14 0G 0.07064 0.02639 0.01595 28 640: 50% 2/4 [00:55<00:55, 27.69s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 10/14 0G 0.06777 0.02661 0.01541 33 640: 75% 3/4 [01:23<00:27, 27.89s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 10/14 0G 0.06365 0.02523 0.01429 27 640: 100% 4/4 [01:54<00:00, 28.50s/it] Class Images Instances P R mAP50 mAP50-95: 0% 0/1 [00:00<?, ?it/s]WARNING ⚠️ NMS time limit 0.900s exceeded Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:03<00:00, 3.89s/it] all 8 8 0.573 0.625 0.721 0.395 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 11/14 0G 0.06045 0.03093 0.01738 43 640: 25% 1/4 [00:27<01:23, 27.87s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 11/14 0G 0.05999 0.02862 0.01447 34 640: 50% 2/4 [00:54<00:53, 27.00s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 11/14 0G 0.06223 0.02631 0.01348 29 640: 75% 3/4 [01:22<00:27, 27.68s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 11/14 0G 0.06261 0.02618 0.01406 36 640: 100% 4/4 [01:48<00:00, 27.24s/it] Class Images Instances P R mAP50 mAP50-95: 0% 0/1 [00:00<?, ?it/s]WARNING ⚠️ NMS time limit 0.900s exceeded Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:04<00:00, 4.93s/it] all 8 8 0.606 0.625 0.666 0.36 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 12/14 0G 0.05826 0.02249 0.00969 28 640: 25% 1/4 [00:26<01:18, 26.27s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 12/14 0G 0.06474 0.0232 0.01208 32 640: 50% 2/4 [00:53<00:53, 26.62s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 12/14 0G 0.06496 0.02332 0.01259 30 640: 75% 3/4 [01:21<00:27, 27.50s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 12/14 0G 0.06266 0.02356 0.01255 33 640: 100% 4/4 [01:49<00:00, 27.28s/it] Class Images Instances P R mAP50 mAP50-95: 0% 0/1 [00:00<?, ?it/s]WARNING ⚠️ NMS time limit 0.900s exceeded Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:03<00:00, 3.84s/it] all 8 8 0.808 0.875 0.92 0.471 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 13/14 0G 0.06822 0.02777 0.01556 40 640: 25% 1/4 [00:27<01:23, 27.89s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 13/14 0G 0.06371 0.02577 0.01287 32 640: 50% 2/4 [00:56<00:56, 28.10s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 13/14 0G 0.06062 0.02516 0.01186 32 640: 75% 3/4 [01:23<00:27, 27.78s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 13/14 0G 0.06319 0.02415 0.01381 28 640: 100% 4/4 [01:55<00:00, 28.95s/it] Class Images Instances P R mAP50 mAP50-95: 0% 0/1 [00:00<?, ?it/s]WARNING ⚠️ NMS time limit 0.900s exceeded Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:03<00:00, 3.84s/it] all 8 8 0.717 0.75 0.863 0.466 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 14/14 0G 0.05416 0.02748 0.009614 39 640: 25% 1/4 [00:27<01:21, 27.17s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 14/14 0G 0.06113 0.02551 0.01132 35 640: 50% 2/4 [00:54<00:54, 27.38s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 14/14 0G 0.0647 0.02451 0.01316 34 640: 75% 3/4 [01:23<00:28, 28.16s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 14/14 0G 0.06033 0.02312 0.01338 25 640: 100% 4/4 [01:52<00:00, 28.20s/it] Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:04<00:00, 4.11s/it] all 8 8 0.88 0.875 0.92 0.576 15 epochs completed in 0.487 hours. Optimizer stripped from runs/train/treino_15_epocas/weights/last.pt, 14.4MB Optimizer stripped from runs/train/treino_15_epocas/weights/best.pt, 14.4MB Validating runs/train/treino_15_epocas/weights/best.pt... Fusing layers... Model summary: 157 layers, 7015519 parameters, 0 gradients, 15.8 GFLOPs Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:03<00:00, 3.61s/it] all 8 8 0.878 0.875 0.92 0.576 bicicleta 8 4 0.833 0.75 0.845 0.514 carro 8 4 0.922 1 0.995 0.638 Results saved to runs/train/treino_15_epocas ✅ Treinamento finalizado: 16:48:29
📊 Análise dos Resultados - 15 Épocas¶
Vamos visualizar as métricas de treinamento e alguns resultados.
In [8]:
# Visualizar gráfico de resultados do treino (15 épocas)
print("📈 Gráfico de métricas de treinamento (15 épocas):")
display(Image(filename='/content/yolov5/runs/train/treino_15_epocas/results.png'))
📈 Gráfico de métricas de treinamento (15 épocas):
In [9]:
# Visualizar exemplos de predições no conjunto de validação
print("🔍 Exemplos de predições no conjunto de validação (15 épocas):")
display(Image(filename='/content/yolov5/runs/train/treino_15_epocas/val_batch0_pred.jpg'))
🔍 Exemplos de predições no conjunto de validação (15 épocas):
In [10]:
# Visualizar matriz de confusão
print("📊 Matriz de Confusão (15 épocas):")
display(Image(filename='/content/yolov5/runs/train/treino_15_epocas/confusion_matrix.png'))
📊 Matriz de Confusão (15 épocas):
In [11]:
# Treinar modelo com 40 épocas
print("🚀 Iniciando treinamento com 40 épocas...")
print(f"⏰ Início: {datetime.now().strftime('%H:%M:%S')}")
!python train.py \
--img 640 \
--batch 16 \
--epochs 40 \
--data dataset.yaml \
--weights yolov5s.pt \
--cache \
--name treino_40_epocas
print(f"✅ Treinamento finalizado: {datetime.now().strftime('%H:%M:%S')}")
🚀 Iniciando treinamento com 40 épocas... ⏰ Início: 16:48:29 wandb: WARNING ⚠️ wandb is deprecated and will be removed in a future release. See supported integrations at https://github.com/ultralytics/yolov5#integrations. 2025-10-20 16:48:43.396880: E external/local_xla/xla/stream_executor/cuda/cuda_fft.cc:467] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered WARNING: All log messages before absl::InitializeLog() is called are written to STDERR E0000 00:00:1760978923.428247 16394 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered E0000 00:00:1760978923.437746 16394 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered W0000 00:00:1760978923.472006 16394 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1760978923.472062 16394 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1760978923.472067 16394 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1760978923.472073 16394 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. wandb: (1) Create a W&B account wandb: (2) Use an existing W&B account wandb: (3) Don't visualize my results wandb: Enter your choice: (30 second timeout) wandb: W&B disabled due to login timeout. train: weights=yolov5s.pt, cfg=, data=dataset.yaml, hyp=data/hyps/hyp.scratch-low.yaml, epochs=40, batch_size=16, imgsz=640, rect=False, resume=False, nosave=False, noval=False, noautoanchor=False, noplots=False, evolve=None, evolve_population=data/hyps, resume_evolve=None, bucket=, cache=ram, image_weights=False, device=, multi_scale=False, single_cls=False, optimizer=SGD, sync_bn=False, workers=8, project=runs/train, name=treino_40_epocas, exist_ok=False, quad=False, cos_lr=False, label_smoothing=0.0, patience=100, freeze=[0], save_period=-1, seed=0, local_rank=-1, entity=None, upload_dataset=False, bbox_interval=-1, artifact_alias=latest, ndjson_console=False, ndjson_file=False github: up to date with https://github.com/ultralytics/yolov5 ✅ YOLOv5 🚀 v7.0-441-g15c0127a Python-3.12.12 torch-2.8.0+cu126 CPU hyperparameters: lr0=0.01, lrf=0.01, momentum=0.937, weight_decay=0.0005, warmup_epochs=3.0, warmup_momentum=0.8, warmup_bias_lr=0.1, box=0.05, cls=0.5, cls_pw=1.0, obj=1.0, obj_pw=1.0, iou_t=0.2, anchor_t=4.0, fl_gamma=0.0, hsv_h=0.015, hsv_s=0.7, hsv_v=0.4, degrees=0.0, translate=0.1, scale=0.5, shear=0.0, perspective=0.0, flipud=0.0, fliplr=0.5, mosaic=1.0, mixup=0.0, copy_paste=0.0 Comet: run 'pip install comet_ml' to automatically track and visualize YOLOv5 🚀 runs in Comet TensorBoard: Start with 'tensorboard --logdir runs/train', view at http://localhost:6006/ Overriding model.yaml nc=80 with nc=2 from n params module arguments 0 -1 1 3520 models.common.Conv [3, 32, 6, 2, 2] 1 -1 1 18560 models.common.Conv [32, 64, 3, 2] 2 -1 1 18816 models.common.C3 [64, 64, 1] 3 -1 1 73984 models.common.Conv [64, 128, 3, 2] 4 -1 2 115712 models.common.C3 [128, 128, 2] 5 -1 1 295424 models.common.Conv [128, 256, 3, 2] 6 -1 3 625152 models.common.C3 [256, 256, 3] 7 -1 1 1180672 models.common.Conv [256, 512, 3, 2] 8 -1 1 1182720 models.common.C3 [512, 512, 1] 9 -1 1 656896 models.common.SPPF [512, 512, 5] 10 -1 1 131584 models.common.Conv [512, 256, 1, 1] 11 -1 1 0 torch.nn.modules.upsampling.Upsample [None, 2, 'nearest'] 12 [-1, 6] 1 0 models.common.Concat [1] 13 -1 1 361984 models.common.C3 [512, 256, 1, False] 14 -1 1 33024 models.common.Conv [256, 128, 1, 1] 15 -1 1 0 torch.nn.modules.upsampling.Upsample [None, 2, 'nearest'] 16 [-1, 4] 1 0 models.common.Concat [1] 17 -1 1 90880 models.common.C3 [256, 128, 1, False] 18 -1 1 147712 models.common.Conv [128, 128, 3, 2] 19 [-1, 14] 1 0 models.common.Concat [1] 20 -1 1 296448 models.common.C3 [256, 256, 1, False] 21 -1 1 590336 models.common.Conv [256, 256, 3, 2] 22 [-1, 10] 1 0 models.common.Concat [1] 23 -1 1 1182720 models.common.C3 [512, 512, 1, False] 24 [17, 20, 23] 1 18879 models.yolo.Detect [2, [[10, 13, 16, 30, 33, 23], [30, 61, 62, 45, 59, 119], [116, 90, 156, 198, 373, 326]], [128, 256, 512]] Model summary: 214 layers, 7025023 parameters, 7025023 gradients, 16.0 GFLOPs Transferred 343/349 items from yolov5s.pt optimizer: SGD(lr=0.01) with parameter groups 57 weight(decay=0.0), 60 weight(decay=0.0005), 60 bias albumentations: 1 validation error for InitSchema size Field required [type=missing, input_value={'scale': (0.8, 1.0), 'ra...: None, 'strict': False}, input_type=dict] For further information visit https://errors.pydantic.dev/2.11/v/missing train: Scanning /content/drive/MyDrive/FIAP/FASE_6/dataset/train/labels.cache... 64 images, 0 backgrounds, 0 corrupt: 100% 64/64 [00:00<?, ?it/s] train: Caching images (0.1GB ram): 100% 64/64 [00:00<00:00, 88.88it/s] val: Scanning /content/drive/MyDrive/FIAP/FASE_6/dataset/valid/labels.cache... 8 images, 0 backgrounds, 0 corrupt: 100% 8/8 [00:00<?, ?it/s] val: Caching images (0.0GB ram): 100% 8/8 [00:00<00:00, 66.97it/s] AutoAnchor: 3.81 anchors/target, 1.000 Best Possible Recall (BPR). Current anchors are a good fit to dataset ✅ Plotting labels to runs/train/treino_40_epocas/labels.jpg... /content/yolov5/train.py:357: FutureWarning: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead. scaler = torch.cuda.amp.GradScaler(enabled=amp) Image sizes 640 train, 640 val Using 2 dataloader workers Logging results to runs/train/treino_40_epocas Starting training for 40 epochs... Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 0/39 0G 0.1256 0.03123 0.03523 34 640: 25% 1/4 [00:45<02:15, 45.28s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 0/39 0G 0.1267 0.03051 0.03385 30 640: 50% 2/4 [01:19<01:17, 38.85s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 0/39 0G 0.1272 0.03027 0.03243 29 640: 75% 3/4 [01:47<00:33, 33.91s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 0/39 0G 0.127 0.02986 0.03252 26 640: 100% 4/4 [02:15<00:00, 33.90s/it] Class Images Instances P R mAP50 mAP50-95: 0% 0/1 [00:00<?, ?it/s]WARNING ⚠️ NMS time limit 0.900s exceeded Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:05<00:00, 5.25s/it] all 8 8 0.00321 0.125 0.00244 0.000968 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 1/39 0G 0.1157 0.02783 0.02899 24 640: 25% 1/4 [00:29<01:29, 29.90s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 1/39 0G 0.1184 0.02846 0.03036 28 640: 50% 2/4 [00:57<00:56, 28.47s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 1/39 0G 0.118 0.02901 0.03123 31 640: 75% 3/4 [01:23<00:27, 27.51s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 1/39 0G 0.1164 0.02926 0.03052 30 640: 100% 4/4 [01:52<00:00, 28.13s/it] Class Images Instances P R mAP50 mAP50-95: 0% 0/1 [00:00<?, ?it/s]WARNING ⚠️ NMS time limit 0.900s exceeded Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:08<00:00, 8.41s/it] all 8 8 0.00323 0.125 0.00248 0.000743 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 2/39 0G 0.1091 0.03049 0.02472 32 640: 25% 1/4 [00:31<01:35, 31.69s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 2/39 0G 0.1035 0.03057 0.02398 31 640: 50% 2/4 [00:57<00:56, 28.50s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 2/39 0G 0.1007 0.03196 0.02358 39 640: 75% 3/4 [01:25<00:28, 28.09s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 2/39 0G 0.09784 0.03159 0.02374 31 640: 100% 4/4 [01:52<00:00, 28.24s/it] Class Images Instances P R mAP50 mAP50-95: 0% 0/1 [00:00<?, ?it/s]WARNING ⚠️ NMS time limit 0.900s exceeded Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:04<00:00, 4.16s/it] all 8 8 0.0036 0.375 0.0112 0.00371 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 3/39 0G 0.09088 0.03291 0.0227 37 640: 25% 1/4 [00:26<01:20, 26.93s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 3/39 0G 0.08113 0.03067 0.01992 24 640: 50% 2/4 [00:54<00:54, 27.07s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 3/39 0G 0.07854 0.02969 0.0221 27 640: 75% 3/4 [01:21<00:27, 27.20s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 3/39 0G 0.08158 0.03034 0.02206 36 640: 100% 4/4 [01:47<00:00, 26.79s/it] Class Images Instances P R mAP50 mAP50-95: 0% 0/1 [00:00<?, ?it/s]WARNING ⚠️ NMS time limit 0.900s exceeded Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:05<00:00, 5.53s/it] all 8 8 0.00358 0.375 0.0699 0.0348 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 4/39 0G 0.09277 0.03064 0.02459 34 640: 25% 1/4 [00:29<01:29, 29.67s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 4/39 0G 0.08859 0.02977 0.02162 30 640: 50% 2/4 [00:56<00:56, 28.27s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 4/39 0G 0.08335 0.03042 0.02008 38 640: 75% 3/4 [01:24<00:28, 28.04s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 4/39 0G 0.08155 0.02954 0.02022 29 640: 100% 4/4 [01:52<00:00, 28.03s/it] Class Images Instances P R mAP50 mAP50-95: 0% 0/1 [00:00<?, ?it/s]WARNING ⚠️ NMS time limit 0.900s exceeded Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:04<00:00, 4.07s/it] all 8 8 0.00356 0.5 0.262 0.144 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 5/39 0G 0.06684 0.03345 0.01782 37 640: 25% 1/4 [00:27<01:22, 27.48s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 5/39 0G 0.07387 0.03117 0.01807 33 640: 50% 2/4 [00:58<00:59, 29.78s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 5/39 0G 0.07506 0.03132 0.01888 41 640: 75% 3/4 [01:27<00:29, 29.28s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 5/39 0G 0.07544 0.03173 0.01852 38 640: 100% 4/4 [01:53<00:00, 28.50s/it] Class Images Instances P R mAP50 mAP50-95: 0% 0/1 [00:00<?, ?it/s]WARNING ⚠️ NMS time limit 0.900s exceeded Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:03<00:00, 3.84s/it] all 8 8 0.00327 0.5 0.295 0.154 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 6/39 0G 0.07764 0.03153 0.01825 39 640: 25% 1/4 [00:26<01:20, 26.79s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 6/39 0G 0.07772 0.03123 0.01653 40 640: 50% 2/4 [00:54<00:54, 27.13s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 6/39 0G 0.07579 0.02939 0.01657 30 640: 75% 3/4 [01:22<00:27, 27.50s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 6/39 0G 0.07241 0.02868 0.01781 30 640: 100% 4/4 [01:49<00:00, 27.30s/it] Class Images Instances P R mAP50 mAP50-95: 0% 0/1 [00:00<?, ?it/s]WARNING ⚠️ NMS time limit 0.900s exceeded Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:04<00:00, 4.95s/it] all 8 8 0.0284 0.625 0.42 0.199 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 7/39 0G 0.06677 0.02995 0.01402 36 640: 25% 1/4 [00:27<01:22, 27.45s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 7/39 0G 0.07024 0.02909 0.01682 39 640: 50% 2/4 [00:55<00:55, 27.65s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 7/39 0G 0.07128 0.02829 0.01671 32 640: 75% 3/4 [01:22<00:27, 27.25s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 7/39 0G 0.07146 0.02813 0.01718 38 640: 100% 4/4 [01:49<00:00, 27.30s/it] Class Images Instances P R mAP50 mAP50-95: 0% 0/1 [00:00<?, ?it/s]WARNING ⚠️ NMS time limit 0.900s exceeded Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:03<00:00, 3.79s/it] all 8 8 0.315 0.375 0.392 0.218 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 8/39 0G 0.07274 0.0296 0.01874 39 640: 25% 1/4 [00:26<01:20, 26.94s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 8/39 0G 0.07232 0.0259 0.01631 30 640: 50% 2/4 [00:54<00:54, 27.31s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 8/39 0G 0.06922 0.02675 0.01567 38 640: 75% 3/4 [01:22<00:27, 27.79s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 8/39 0G 0.06897 0.02596 0.0149 31 640: 100% 4/4 [01:50<00:00, 27.68s/it] Class Images Instances P R mAP50 mAP50-95: 0% 0/1 [00:00<?, ?it/s]WARNING ⚠️ NMS time limit 0.900s exceeded Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:03<00:00, 3.76s/it] all 8 8 0.774 0.506 0.654 0.404 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 9/39 0G 0.06048 0.02548 0.01414 35 640: 25% 1/4 [00:26<01:20, 26.76s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 9/39 0G 0.06044 0.02807 0.01212 44 640: 50% 2/4 [00:54<00:54, 27.45s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 9/39 0G 0.05967 0.02729 0.01307 35 640: 75% 3/4 [01:21<00:27, 27.20s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 9/39 0G 0.06029 0.02532 0.01251 27 640: 100% 4/4 [01:48<00:00, 27.01s/it] Class Images Instances P R mAP50 mAP50-95: 0% 0/1 [00:00<?, ?it/s]WARNING ⚠️ NMS time limit 0.900s exceeded Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:03<00:00, 3.96s/it] all 8 8 0.441 0.625 0.549 0.283 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 10/39 0G 0.06632 0.0276 0.01184 39 640: 25% 1/4 [00:27<01:21, 27.20s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 10/39 0G 0.0707 0.0237 0.01233 28 640: 50% 2/4 [00:54<00:54, 27.45s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 10/39 0G 0.06791 0.02375 0.01207 33 640: 75% 3/4 [01:26<00:29, 29.29s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 10/39 0G 0.06457 0.02228 0.01134 27 640: 100% 4/4 [01:53<00:00, 28.42s/it] Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:03<00:00, 3.65s/it] all 8 8 0.898 0.75 0.854 0.46 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 11/39 0G 0.06159 0.02662 0.01439 43 640: 25% 1/4 [00:27<01:23, 27.87s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 11/39 0G 0.05935 0.02505 0.01163 34 640: 50% 2/4 [00:54<00:54, 27.27s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 11/39 0G 0.06062 0.02324 0.0108 29 640: 75% 3/4 [01:23<00:27, 27.90s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 11/39 0G 0.06165 0.0231 0.01128 36 640: 100% 4/4 [01:49<00:00, 27.47s/it] Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:04<00:00, 4.66s/it] all 8 8 0.825 0.822 0.858 0.369 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 12/39 0G 0.05951 0.0195 0.007279 28 640: 25% 1/4 [00:26<01:20, 26.86s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 12/39 0G 0.06709 0.01976 0.009518 32 640: 50% 2/4 [00:53<00:53, 26.54s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 12/39 0G 0.06584 0.02001 0.01011 30 640: 75% 3/4 [01:22<00:27, 27.95s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 12/39 0G 0.06411 0.01999 0.01011 33 640: 100% 4/4 [01:49<00:00, 27.46s/it] Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:03<00:00, 3.35s/it] all 8 8 0.923 0.875 0.915 0.579 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 13/39 0G 0.06866 0.02378 0.01207 40 640: 25% 1/4 [00:26<01:19, 26.53s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 13/39 0G 0.06444 0.02149 0.01014 32 640: 50% 2/4 [00:55<00:55, 27.89s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 13/39 0G 0.06425 0.02093 0.009063 32 640: 75% 3/4 [01:21<00:27, 27.28s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 13/39 0G 0.06791 0.01996 0.01085 28 640: 100% 4/4 [01:48<00:00, 27.09s/it] Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:04<00:00, 4.40s/it] all 8 8 0.169 0.894 0.228 0.121 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 14/39 0G 0.05983 0.02323 0.00673 39 640: 25% 1/4 [00:25<01:16, 25.49s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 14/39 0G 0.0623 0.02146 0.00849 35 640: 50% 2/4 [00:51<00:51, 25.70s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 14/39 0G 0.06299 0.02071 0.009687 34 640: 75% 3/4 [01:18<00:26, 26.54s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 14/39 0G 0.05861 0.01949 0.009315 25 640: 100% 4/4 [01:46<00:00, 26.69s/it] Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:03<00:00, 3.22s/it] all 8 8 0.687 1 0.808 0.442 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 15/39 0G 0.05434 0.01905 0.005311 33 640: 25% 1/4 [00:26<01:19, 26.42s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 15/39 0G 0.05827 0.0188 0.005822 34 640: 50% 2/4 [00:54<00:55, 27.57s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 15/39 0G 0.05913 0.01989 0.007395 38 640: 75% 3/4 [01:21<00:27, 27.32s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 15/39 0G 0.05776 0.02049 0.006868 41 640: 100% 4/4 [01:48<00:00, 27.04s/it] Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:03<00:00, 3.67s/it] all 8 8 0.645 0.875 0.922 0.423 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 16/39 0G 0.06076 0.01779 0.007256 32 640: 25% 1/4 [00:30<01:31, 30.43s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 16/39 0G 0.06141 0.01892 0.007303 38 640: 50% 2/4 [00:58<00:58, 29.30s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 16/39 0G 0.0659 0.01755 0.009285 28 640: 75% 3/4 [01:26<00:28, 28.40s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 16/39 0G 0.06441 0.01808 0.008683 38 640: 100% 4/4 [01:54<00:00, 28.57s/it] Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:03<00:00, 3.16s/it] all 8 8 0.695 1 0.87 0.458 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 17/39 0G 0.0587 0.01957 0.005002 36 640: 25% 1/4 [00:26<01:19, 26.65s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 17/39 0G 0.05637 0.01879 0.004963 36 640: 50% 2/4 [00:53<00:53, 26.63s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 17/39 0G 0.05936 0.01829 0.00589 33 640: 75% 3/4 [01:20<00:26, 26.78s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 17/39 0G 0.06001 0.01737 0.006168 25 640: 100% 4/4 [01:47<00:00, 26.99s/it] Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:03<00:00, 3.58s/it] all 8 8 0.973 0.867 0.941 0.337 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 18/39 0G 0.05839 0.01863 0.007206 36 640: 25% 1/4 [00:26<01:19, 26.38s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 18/39 0G 0.06197 0.0191 0.007979 41 640: 50% 2/4 [00:55<00:55, 27.75s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 18/39 0G 0.06264 0.0184 0.006928 34 640: 75% 3/4 [01:23<00:28, 28.20s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 18/39 0G 0.0635 0.01793 0.007017 31 640: 100% 4/4 [01:50<00:00, 27.70s/it] Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:03<00:00, 3.05s/it] all 8 8 0.792 0.619 0.79 0.276 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 19/39 0G 0.07101 0.01465 0.004276 30 640: 25% 1/4 [00:27<01:21, 27.22s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 19/39 0G 0.06948 0.01427 0.00418 24 640: 50% 2/4 [00:56<00:56, 28.24s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 19/39 0G 0.06523 0.0161 0.005904 41 640: 75% 3/4 [01:22<00:27, 27.27s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 19/39 0G 0.06273 0.01586 0.006118 28 640: 100% 4/4 [01:50<00:00, 27.67s/it] Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:04<00:00, 4.57s/it] all 8 8 0.89 0.838 0.953 0.688 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 20/39 0G 0.05421 0.0177 0.004143 33 640: 25% 1/4 [00:26<01:19, 26.49s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 20/39 0G 0.05569 0.01899 0.004246 40 640: 50% 2/4 [00:53<00:53, 26.66s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 20/39 0G 0.05754 0.01861 0.005261 36 640: 75% 3/4 [01:20<00:27, 27.06s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 20/39 0G 0.05183 0.01778 0.004818 31 640: 100% 4/4 [01:49<00:00, 27.26s/it] Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:03<00:00, 3.02s/it] all 8 8 0.864 0.865 0.916 0.644 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 21/39 0G 0.05756 0.01657 0.004643 32 640: 25% 1/4 [00:27<01:23, 27.89s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 21/39 0G 0.05919 0.01564 0.003861 31 640: 50% 2/4 [00:55<00:55, 27.99s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 21/39 0G 0.05986 0.01612 0.004192 37 640: 75% 3/4 [01:25<00:28, 28.88s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 21/39 0G 0.06057 0.01611 0.004262 35 640: 100% 4/4 [01:52<00:00, 28.21s/it] Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:03<00:00, 3.43s/it] all 8 8 0.535 0.75 0.708 0.497 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 22/39 0G 0.06287 0.01575 0.003479 34 640: 25% 1/4 [00:26<01:20, 26.76s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 22/39 0G 0.05975 0.01621 0.00375 37 640: 50% 2/4 [00:53<00:53, 26.60s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 22/39 0G 0.05781 0.01505 0.003552 26 640: 75% 3/4 [01:20<00:27, 27.08s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 22/39 0G 0.05823 0.01496 0.004222 27 640: 100% 4/4 [01:49<00:00, 27.32s/it] Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:02<00:00, 2.97s/it] all 8 8 0.939 0.875 0.941 0.62 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 23/39 0G 0.05566 0.01995 0.007277 40 640: 25% 1/4 [00:27<01:21, 27.13s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 23/39 0G 0.05664 0.01903 0.006923 38 640: 50% 2/4 [00:55<00:55, 27.82s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 23/39 0G 0.05165 0.01766 0.005838 29 640: 75% 3/4 [01:23<00:27, 27.78s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 23/39 0G 0.05308 0.01699 0.005561 29 640: 100% 4/4 [01:51<00:00, 27.80s/it] Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:03<00:00, 3.00s/it] all 8 8 0.617 0.75 0.702 0.445 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 24/39 0G 0.0541 0.01371 0.002871 30 640: 25% 1/4 [00:26<01:20, 26.88s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 24/39 0G 0.05703 0.01542 0.003299 36 640: 50% 2/4 [00:54<00:54, 27.36s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 24/39 0G 0.05807 0.01644 0.004296 41 640: 75% 3/4 [01:21<00:27, 27.26s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 24/39 0G 0.05958 0.0159 0.004114 33 640: 100% 4/4 [01:49<00:00, 27.50s/it] Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:03<00:00, 3.38s/it] all 8 8 0.903 0.875 0.933 0.538 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 25/39 0G 0.0522 0.01511 0.002875 33 640: 25% 1/4 [00:27<01:21, 27.11s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 25/39 0G 0.05449 0.01618 0.004479 34 640: 50% 2/4 [00:55<00:55, 27.87s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 25/39 0G 0.0545 0.01662 0.004615 36 640: 75% 3/4 [01:23<00:27, 27.93s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 25/39 0G 0.05397 0.0163 0.004767 30 640: 100% 4/4 [01:50<00:00, 27.71s/it] Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:02<00:00, 2.93s/it] all 8 8 0.728 0.875 0.857 0.586 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 26/39 0G 0.05362 0.01849 0.004678 37 640: 25% 1/4 [00:27<01:22, 27.52s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 26/39 0G 0.05234 0.01811 0.00568 37 640: 50% 2/4 [00:55<00:55, 27.90s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 26/39 0G 0.05309 0.01763 0.005449 34 640: 75% 3/4 [01:22<00:27, 27.55s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 26/39 0G 0.0516 0.01665 0.004941 30 640: 100% 4/4 [01:49<00:00, 27.45s/it] Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:04<00:00, 4.02s/it] all 8 8 0.905 0.875 0.926 0.59 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 27/39 0G 0.04994 0.01323 0.00385 31 640: 25% 1/4 [00:33<01:41, 33.78s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 27/39 0G 0.05481 0.01383 0.004324 32 640: 50% 2/4 [01:01<01:00, 30.25s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 27/39 0G 0.05288 0.01399 0.003784 31 640: 75% 3/4 [01:28<00:28, 28.93s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 27/39 0G 0.05281 0.01466 0.0036 37 640: 100% 4/4 [01:56<00:00, 29.06s/it] Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:02<00:00, 2.90s/it] all 8 8 0.863 0.875 0.941 0.681 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 28/39 0G 0.04644 0.01408 0.002826 28 640: 25% 1/4 [00:26<01:20, 26.69s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 28/39 0G 0.04648 0.01471 0.003315 35 640: 50% 2/4 [00:57<00:58, 29.26s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 28/39 0G 0.04743 0.01586 0.004018 36 640: 75% 3/4 [01:24<00:28, 28.21s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 28/39 0G 0.04765 0.01533 0.005307 30 640: 100% 4/4 [01:50<00:00, 27.71s/it] Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:03<00:00, 3.95s/it] all 8 8 0.892 0.875 0.97 0.746 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 29/39 0G 0.04652 0.01876 0.004169 39 640: 25% 1/4 [00:25<01:17, 25.74s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 29/39 0G 0.04956 0.01686 0.00422 30 640: 50% 2/4 [00:52<00:52, 26.36s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 29/39 0G 0.04678 0.01661 0.003628 31 640: 75% 3/4 [01:20<00:26, 26.93s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 29/39 0G 0.0443 0.01569 0.003496 26 640: 100% 4/4 [01:47<00:00, 26.83s/it] Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:02<00:00, 2.96s/it] all 8 8 0.814 0.875 0.953 0.486 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 30/39 0G 0.05677 0.01579 0.00356 37 640: 25% 1/4 [00:26<01:20, 26.77s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 30/39 0G 0.05387 0.01481 0.003214 31 640: 50% 2/4 [00:54<00:54, 27.05s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 30/39 0G 0.05184 0.01443 0.002716 31 640: 75% 3/4 [01:20<00:26, 26.86s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 30/39 0G 0.05201 0.01546 0.003304 39 640: 100% 4/4 [01:47<00:00, 26.94s/it] Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:02<00:00, 2.94s/it] all 8 8 0.904 0.893 0.995 0.755 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 31/39 0G 0.0481 0.01726 0.004034 38 640: 25% 1/4 [00:26<01:20, 26.70s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 31/39 0G 0.0475 0.01818 0.003907 40 640: 50% 2/4 [00:55<00:56, 28.03s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 31/39 0G 0.04591 0.01758 0.003678 35 640: 75% 3/4 [01:23<00:27, 27.76s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 31/39 0G 0.04535 0.01645 0.003507 28 640: 100% 4/4 [01:50<00:00, 27.58s/it] Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:03<00:00, 3.19s/it] all 8 8 0.988 0.897 0.995 0.715 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 32/39 0G 0.04321 0.01355 0.003011 33 640: 25% 1/4 [00:25<01:17, 25.82s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 32/39 0G 0.04408 0.01333 0.003337 33 640: 50% 2/4 [00:56<00:57, 28.87s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 32/39 0G 0.04244 0.0145 0.002983 37 640: 75% 3/4 [01:23<00:28, 28.09s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 32/39 0G 0.04521 0.01568 0.003275 42 640: 100% 4/4 [01:51<00:00, 27.87s/it] Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:02<00:00, 2.93s/it] all 8 8 0.867 0.841 0.941 0.701 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 33/39 0G 0.04022 0.01716 0.00217 38 640: 25% 1/4 [00:26<01:19, 26.48s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 33/39 0G 0.04293 0.0178 0.002888 40 640: 50% 2/4 [00:54<00:54, 27.36s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 33/39 0G 0.04611 0.01624 0.004394 33 640: 75% 3/4 [01:22<00:27, 27.51s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 33/39 0G 0.04728 0.01624 0.004526 40 640: 100% 4/4 [01:50<00:00, 27.68s/it] Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:02<00:00, 2.89s/it] all 8 8 0.956 0.983 0.995 0.708 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 34/39 0G 0.05041 0.01402 0.01012 32 640: 25% 1/4 [00:26<01:18, 26.11s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 34/39 0G 0.04607 0.0155 0.006527 36 640: 50% 2/4 [00:51<00:51, 25.90s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 34/39 0G 0.04331 0.01495 0.005127 31 640: 75% 3/4 [01:19<00:26, 26.64s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 34/39 0G 0.04343 0.01578 0.004861 40 640: 100% 4/4 [01:47<00:00, 26.79s/it] Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:02<00:00, 2.88s/it] all 8 8 0.966 0.981 0.995 0.732 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 35/39 0G 0.04079 0.01288 0.001574 30 640: 25% 1/4 [00:26<01:19, 26.64s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 35/39 0G 0.04164 0.01371 0.002073 34 640: 50% 2/4 [00:54<00:54, 27.36s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 35/39 0G 0.03978 0.01438 0.002213 37 640: 75% 3/4 [01:20<00:26, 26.82s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 35/39 0G 0.03897 0.01411 0.002119 34 640: 100% 4/4 [01:48<00:00, 27.05s/it] Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:04<00:00, 4.43s/it] all 8 8 0.977 0.983 0.995 0.798 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 36/39 0G 0.04518 0.01376 0.005489 32 640: 25% 1/4 [00:25<01:16, 25.45s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 36/39 0G 0.04377 0.01326 0.004206 28 640: 50% 2/4 [00:53<00:53, 26.89s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 36/39 0G 0.04344 0.01342 0.003941 31 640: 75% 3/4 [01:20<00:27, 27.22s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 36/39 0G 0.04264 0.01333 0.003694 26 640: 100% 4/4 [01:48<00:00, 27.17s/it] Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:08<00:00, 8.37s/it] all 8 8 0.983 0.987 0.995 0.76 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 37/39 0G 0.03653 0.0166 0.002384 38 640: 25% 1/4 [00:26<01:19, 26.58s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 37/39 0G 0.03905 0.01406 0.00339 28 640: 50% 2/4 [00:54<00:54, 27.26s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 37/39 0G 0.04223 0.01455 0.003688 38 640: 75% 3/4 [01:22<00:27, 27.53s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 37/39 0G 0.04019 0.01436 0.003357 31 640: 100% 4/4 [01:49<00:00, 27.37s/it] Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:02<00:00, 3.00s/it] all 8 8 0.985 0.995 0.995 0.745 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 38/39 0G 0.04213 0.012 0.002828 25 640: 25% 1/4 [00:25<01:17, 25.76s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 38/39 0G 0.03804 0.01334 0.003039 34 640: 50% 2/4 [00:51<00:51, 25.79s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 38/39 0G 0.0385 0.01489 0.00336 44 640: 75% 3/4 [01:18<00:26, 26.20s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 38/39 0G 0.03682 0.0139 0.00309 22 640: 100% 4/4 [01:45<00:00, 26.32s/it] Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:02<00:00, 2.91s/it] all 8 8 0.985 0.996 0.995 0.725 Epoch GPU_mem box_loss obj_loss cls_loss Instances Size 0% 0/4 [00:00<?, ?it/s]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 39/39 0G 0.04925 0.01303 0.005979 33 640: 25% 1/4 [00:26<01:19, 26.63s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 39/39 0G 0.04677 0.0143 0.005792 31 640: 50% 2/4 [00:53<00:53, 26.86s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 39/39 0G 0.04606 0.01506 0.004959 41 640: 75% 3/4 [01:20<00:26, 26.88s/it]/content/yolov5/train.py:414: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. with torch.cuda.amp.autocast(amp): 39/39 0G 0.04353 0.01463 0.004131 34 640: 100% 4/4 [01:48<00:00, 27.00s/it] Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:02<00:00, 2.91s/it] all 8 8 0.983 0.991 0.995 0.749 40 epochs completed in 1.279 hours. Optimizer stripped from runs/train/treino_40_epocas/weights/last.pt, 14.4MB Optimizer stripped from runs/train/treino_40_epocas/weights/best.pt, 14.4MB Validating runs/train/treino_40_epocas/weights/best.pt... Fusing layers... Model summary: 157 layers, 7015519 parameters, 0 gradients, 15.8 GFLOPs Class Images Instances P R mAP50 mAP50-95: 100% 1/1 [00:03<00:00, 3.97s/it] all 8 8 0.976 0.983 0.995 0.791 bicicleta 8 4 1 0.966 0.995 0.908 carro 8 4 0.952 1 0.995 0.673 Results saved to runs/train/treino_40_epocas ✅ Treinamento finalizado: 18:06:25
📊 Análise dos Resultados - 40 Épocas¶
In [12]:
# Visualizar gráfico de resultados do treino (40 épocas)
print("📈 Gráfico de métricas de treinamento (40 épocas):")
display(Image(filename='/content/yolov5/runs/train/treino_40_epocas/results.png'))
📈 Gráfico de métricas de treinamento (40 épocas):
In [13]:
# Visualizar exemplos de predições no conjunto de validação
print("🔍 Exemplos de predições no conjunto de validação (40 épocas):")
display(Image(filename='/content/yolov5/runs/train/treino_40_epocas/val_batch0_pred.jpg'))
🔍 Exemplos de predições no conjunto de validação (40 épocas):
In [14]:
# Visualizar matriz de confusão
print("📊 Matriz de Confusão (40 épocas):")
display(Image(filename='/content/yolov5/runs/train/treino_40_epocas/confusion_matrix.png'))
📊 Matriz de Confusão (40 épocas):
📊 6. COMPARAÇÃO ENTRE 15 vs 40 ÉPOCAS¶
Vamos analisar comparativamente os resultados dos dois treinamentos.
In [33]:
# Comparação visual lado a lado
from IPython.display import Image, display
import matplotlib.pyplot as plt
from PIL import Image as PILImage
import numpy as np
print("🔄 Comparação Visual dos Gráficos de Resultados:\n")
# Carregar as imagens
img1 = PILImage.open('/content/yolov5/runs/train/treino_15_epocas/results.png')
img2 = PILImage.open('/content/yolov5/runs/train/treino_40_epocas/results.png')
# Criar figura com 2 subplots lado a lado
fig, axes = plt.subplots(1, 2, figsize=(20, 8))
# Mostrar imagens
axes[0].imshow(img1)
axes[0].set_title('15 Épocas', fontsize=16, fontweight='bold')
axes[0].axis('off')
axes[1].imshow(img2)
axes[1].set_title('40 Épocas', fontsize=16, fontweight='bold')
axes[1].axis('off')
plt.tight_layout()
plt.show()
🔄 Comparação Visual dos Gráficos de Resultados:
📊 COMPARAÇÃO AUTOMÁTICA ENTRE TREINAMENTOS¶
Este código lê automaticamente os resultados reais dos treinamentos e cria tabelas e análises comparativas.
In [34]:
import pandas as pd
import os
import re
from datetime import datetime
def extrair_metricas_completas(caminho_treino):
"""
Extrai métricas completas do arquivo results.csv gerado pelo YOLOv5
Args:
caminho_treino: Caminho para a pasta do treinamento (ex: 'runs/train/treino_15_epocas')
Returns:
dict com as métricas da última época
"""
results_csv = f'{caminho_treino}/results.csv'
if not os.path.exists(results_csv):
print(f"❌ Arquivo não encontrado: {results_csv}")
return None
try:
# Ler CSV
df = pd.read_csv(results_csv)
# Limpar nomes das colunas (remover espaços)
df.columns = df.columns.str.strip()
# Pegar última linha (última época)
ultima_epoca = df.iloc[-1]
# Extrair métricas
metricas = {
'epoca': int(ultima_epoca['epoch']) + 1, # +1 porque começa do 0
'box_loss': float(ultima_epoca['train/box_loss']),
'obj_loss': float(ultima_epoca['train/obj_loss']),
'cls_loss': float(ultima_epoca['train/cls_loss']),
'precision': float(ultima_epoca['metrics/precision']),
'recall': float(ultima_epoca['metrics/recall']),
'mAP50': float(ultima_epoca['metrics/mAP_0.5']),
'mAP50_95': float(ultima_epoca['metrics/mAP_0.5:0.95']),
'val_box_loss': float(ultima_epoca['val/box_loss']),
'val_obj_loss': float(ultima_epoca['val/obj_loss']),
'val_cls_loss': float(ultima_epoca['val/cls_loss'])
}
return metricas
except Exception as e:
print(f"❌ Erro ao processar {results_csv}: {e}")
return None
# Listar treinamentos disponíveis
train_dir = '/content/yolov5/runs/train'
if os.path.exists(train_dir):
treinamentos = [d for d in os.listdir(train_dir)
if os.path.isdir(os.path.join(train_dir, d))]
print("📂 Treinamentos encontrados:")
for t in treinamentos:
print(f" - {t}")
else:
print("❌ Pasta de treinamentos não encontrada!")
treinamentos = []
# Definir quais treinamentos comparar
# Aqui a gente pode incluir os treinamentos disponíveis para comparação
treinos_para_comparar = ['treino_15_epocas', 'treino_40_epocas']
print("\n🔍 Extraindo métricas dos treinamentos...\n")
resultados = {}
for treino in treinos_para_comparar:
caminho = f'/content/yolov5/runs/train/{treino}'
if os.path.exists(caminho):
print(f"📊 Processando: {treino}")
metricas = extrair_metricas_completas(caminho)
if metricas:
resultados[treino] = metricas
print(f" ✅ {metricas['epoca']} épocas completadas")
print(f" 📈 mAP@0.5: {metricas['mAP50']:.4f}")
print(f" 📈 mAP@0.5:0.95: {metricas['mAP50_95']:.4f}\n")
else:
print(f" ❌ Falha ao extrair métricas\n")
else:
print(f"❌ Pasta não encontrada: {treino}\n")
# Criar DataFrame comparativo
if len(resultados) >= 2:
print("=" * 80)
print("📊 TABELA COMPARATIVA AUTOMÁTICA")
print("=" * 80 + "\n")
# Preparar dados para o DataFrame
dados_comparacao = {
'Métrica': [],
treinos_para_comparar[0]: [],
treinos_para_comparar[1]: [],
'Diferença': [],
'Variação %': []
}
# Lista de métricas a comparar
metricas_para_mostrar = [
('mAP50', 'mAP@0.5 (Geral)', 'melhora'),
('mAP50_95', 'mAP@0.5:0.95 (Geral)', 'melhora'),
('precision', 'Precision (Geral)', 'melhora'),
('recall', 'Recall (Geral)', 'melhora'),
('box_loss', 'Box Loss (Train)', 'reduz'),
('obj_loss', 'Object Loss (Train)', 'reduz'),
('cls_loss', 'Class Loss (Train)', 'reduz'),
('val_box_loss', 'Box Loss (Val)', 'reduz'),
('val_obj_loss', 'Object Loss (Val)', 'reduz'),
('val_cls_loss', 'Class Loss (Val)', 'reduz')
]
treino1_nome = treinos_para_comparar[0]
treino2_nome = treinos_para_comparar[1]
treino1 = resultados[treino1_nome]
treino2 = resultados[treino2_nome]
for metrica_key, metrica_nome, tipo in metricas_para_mostrar:
valor1 = treino1[metrica_key]
valor2 = treino2[metrica_key]
diferenca = valor2 - valor1
# Calcular variação percentual
if valor1 != 0:
variacao_pct = (diferenca / abs(valor1)) * 100
else:
variacao_pct = 0
# Determinar símbolo
if tipo == 'melhora':
simbolo = '↑' if diferenca > 0 else '↓'
cor_emoji = '🟢' if diferenca > 0 else '🔴'
else: # tipo == 'reduz'
simbolo = '↓' if diferenca < 0 else '↑'
cor_emoji = '🟢' if diferenca < 0 else '🔴'
dados_comparacao['Métrica'].append(metrica_nome)
dados_comparacao[treino1_nome].append(f"{valor1:.4f}")
dados_comparacao[treino2_nome].append(f"{valor2:.4f}")
dados_comparacao['Diferença'].append(f"{simbolo} {abs(diferenca):.4f}")
dados_comparacao['Variação %'].append(f"{cor_emoji} {variacao_pct:+.2f}%")
# Adicionar informação de épocas e tempo
dados_comparacao['Métrica'].append('─' * 30)
dados_comparacao[treino1_nome].append('─' * 10)
dados_comparacao[treino2_nome].append('─' * 10)
dados_comparacao['Diferença'].append('─' * 10)
dados_comparacao['Variação %'].append('─' * 10)
dados_comparacao['Métrica'].append('Épocas Treinadas')
dados_comparacao[treino1_nome].append(f"{treino1['epoca']}")
dados_comparacao[treino2_nome].append(f"{treino2['epoca']}")
dados_comparacao['Diferença'].append(f"+{treino2['epoca'] - treino1['epoca']}")
dados_comparacao['Variação %'].append(f"{((treino2['epoca'] / treino1['epoca']) - 1) * 100:+.1f}%")
# Criar DataFrame
df_comparacao = pd.DataFrame(dados_comparacao)
# Exibir tabela
print(df_comparacao.to_string(index=False))
print("\n" + "=" * 80)
else:
print("⚠️ Não há treinamentos suficientes para comparar")
📂 Treinamentos encontrados:
- treino_15_epocas
- treino_40_epocas
🔍 Extraindo métricas dos treinamentos...
📊 Processando: treino_15_epocas
✅ 15 épocas completadas
📈 mAP@0.5: 0.9200
📈 mAP@0.5:0.95: 0.5759
📊 Processando: treino_40_epocas
✅ 40 épocas completadas
📈 mAP@0.5: 0.9950
📈 mAP@0.5:0.95: 0.7493
================================================================================
📊 TABELA COMPARATIVA AUTOMÁTICA
================================================================================
Métrica treino_15_epocas treino_40_epocas Diferença Variação %
mAP@0.5 (Geral) 0.9200 0.9950 ↑ 0.0750 🟢 +8.15%
mAP@0.5:0.95 (Geral) 0.5759 0.7493 ↑ 0.1735 🟢 +30.12%
Precision (Geral) 0.8804 0.9830 ↑ 0.1025 🟢 +11.65%
Recall (Geral) 0.8750 0.9908 ↑ 0.1158 🟢 +13.23%
Box Loss (Train) 0.0603 0.0435 ↓ 0.0168 🟢 -27.86%
Object Loss (Train) 0.0231 0.0146 ↓ 0.0085 🟢 -36.74%
Class Loss (Train) 0.0134 0.0041 ↓ 0.0092 🟢 -69.12%
Box Loss (Val) 0.0294 0.0184 ↓ 0.0110 🟢 -37.46%
Object Loss (Val) 0.0166 0.0074 ↓ 0.0092 🟢 -55.57%
Class Loss (Val) 0.0055 0.0011 ↓ 0.0044 🟢 -80.21%
────────────────────────────── ────────── ────────── ────────── ──────────
Épocas Treinadas 15 40 +25 +166.7%
================================================================================
In [35]:
# Análise visual com gráfico de barras
if len(resultados) >= 2:
import matplotlib.pyplot as plt
import numpy as np
print("\n📊 VISUALIZAÇÃO COMPARATIVA\n")
# Métricas principais para visualizar
metricas_viz = ['mAP50', 'mAP50_95', 'precision', 'recall']
nomes_viz = ['mAP@0.5', 'mAP@0.5:0.95', 'Precision', 'Recall']
treino1 = resultados[treinos_para_comparar[0]]
treino2 = resultados[treinos_para_comparar[1]]
# Extrair valores
valores_treino1 = [treino1[m] for m in metricas_viz]
valores_treino2 = [treino2[m] for m in metricas_viz]
# Configurar gráfico
x = np.arange(len(nomes_viz))
width = 0.35
fig, ax = plt.subplots(figsize=(12, 6))
bars1 = ax.bar(x - width/2, valores_treino1, width,
label=f'{treinos_para_comparar[0]} ({treino1["epoca"]} épocas)',
color='#3498db', alpha=0.8)
bars2 = ax.bar(x + width/2, valores_treino2, width,
label=f'{treinos_para_comparar[1]} ({treino2["epoca"]} épocas)',
color='#2ecc71', alpha=0.8)
# Adicionar valores nas barras
for bars in [bars1, bars2]:
for bar in bars:
height = bar.get_height()
ax.text(bar.get_x() + bar.get_width()/2., height,
f'{height:.3f}',
ha='center', va='bottom', fontsize=10, fontweight='bold')
# Configurações do gráfico
ax.set_xlabel('Métricas', fontsize=12, fontweight='bold')
ax.set_ylabel('Valores', fontsize=12, fontweight='bold')
ax.set_title('Comparação de Métricas entre Treinamentos',
fontsize=14, fontweight='bold', pad=20)
ax.set_xticks(x)
ax.set_xticklabels(nomes_viz)
ax.legend(fontsize=11)
ax.grid(axis='y', alpha=0.3, linestyle='--')
ax.set_ylim([0, 1.1])
plt.tight_layout()
plt.show()
📊 VISUALIZAÇÃO COMPARATIVA
In [36]:
# Resumo executivo
if len(resultados) >= 2:
print("\n" + "=" * 80)
print("📝 RESUMO EXECUTIVO DA COMPARAÇÃO")
print("=" * 80 + "\n")
treino1 = resultados[treinos_para_comparar[0]]
treino2 = resultados[treinos_para_comparar[1]]
# Calcular melhorias
melhoria_mAP50 = ((treino2['mAP50'] - treino1['mAP50']) / treino1['mAP50']) * 100
melhoria_mAP50_95 = ((treino2['mAP50_95'] - treino1['mAP50_95']) / treino1['mAP50_95']) * 100
melhoria_precision = ((treino2['precision'] - treino1['precision']) / treino1['precision']) * 100
melhoria_recall = ((treino2['recall'] - treino1['recall']) / treino1['recall']) * 100
reducao_loss = ((treino1['val_box_loss'] - treino2['val_box_loss']) / treino1['val_box_loss']) * 100
print(f"🎯 **Configurações:**")
print(f" • Treino 1: {treino1['epoca']} épocas")
print(f" • Treino 2: {treino2['epoca']} épocas")
print(f" • Diferença: +{treino2['epoca'] - treino1['epoca']} épocas ({((treino2['epoca'] / treino1['epoca']) - 1) * 100:.1f}% mais treinamento)\n")
print(f"📈 **Melhorias em Performance:**")
print(f" • mAP@0.5: {treino1['mAP50']:.4f} → {treino2['mAP50']:.4f} ({melhoria_mAP50:+.2f}%)")
print(f" • mAP@0.5:0.95: {treino1['mAP50_95']:.4f} → {treino2['mAP50_95']:.4f} ({melhoria_mAP50_95:+.2f}%)")
print(f" • Precision: {treino1['precision']:.4f} → {treino2['precision']:.4f} ({melhoria_precision:+.2f}%)")
print(f" • Recall: {treino1['recall']:.4f} → {treino2['recall']:.4f} ({melhoria_recall:+.2f}%)\n")
print(f"📉 **Redução de Loss (Erro):**")
print(f" • Validation Box Loss: {treino1['val_box_loss']:.4f} → {treino2['val_box_loss']:.4f} (-{reducao_loss:.2f}%)\n")
print(f"🏆 **Conclusão:**")
if melhoria_mAP50 > 5:
print(f" ✅ Melhoria SIGNIFICATIVA no mAP@0.5 (+{melhoria_mAP50:.1f}%)")
print(f" ✅ O aumento de épocas trouxe ganhos expressivos de performance")
elif melhoria_mAP50 > 0:
print(f" ✅ Melhoria MODERADA no mAP@0.5 (+{melhoria_mAP50:.1f}%)")
print(f" ⚠️ Ganhos incrementais - avaliar custo-benefício do tempo extra")
else:
print(f" ⚠️ Sem melhoria significativa no mAP@0.5")
print(f" 💡 Possível saturação do modelo - considere outras otimizações")
if melhoria_mAP50_95 > 10:
print(f" ✅ Melhoria EXPRESSIVA no mAP@0.5:0.95 (+{melhoria_mAP50_95:.1f}%)")
print(f" ✅ Modelo ficou mais robusto em diferentes thresholds de IoU")
print("\n" + "=" * 80)
================================================================================ 📝 RESUMO EXECUTIVO DA COMPARAÇÃO ================================================================================ 🎯 **Configurações:** • Treino 1: 15 épocas • Treino 2: 40 épocas • Diferença: +25 épocas (166.7% mais treinamento) 📈 **Melhorias em Performance:** • mAP@0.5: 0.9200 → 0.9950 (+8.15%) • mAP@0.5:0.95: 0.5759 → 0.7493 (+30.12%) • Precision: 0.8804 → 0.9830 (+11.65%) • Recall: 0.8750 → 0.9908 (+13.23%) 📉 **Redução de Loss (Erro):** • Validation Box Loss: 0.0294 → 0.0184 (-37.46%) 🏆 **Conclusão:** ✅ Melhoria SIGNIFICATIVA no mAP@0.5 (+8.2%) ✅ O aumento de épocas trouxe ganhos expressivos de performance ✅ Melhoria EXPRESSIVA no mAP@0.5:0.95 (+30.1%) ✅ Modelo ficou mais robusto em diferentes thresholds de IoU ================================================================================
🎯 Análise Crítica: 15 vs 40 Épocas¶
1. Evolução Geral do Modelo¶
Com o aumento de 15 para 40 épocas, observamos melhorias significativas:
- mAP@0.5: Aumentou de 92% para 99.5% (+7.5 pontos percentuais)
- mAP@0.5:0.95: Salto expressivo de 57.6% para 79.1% (+21.5 pontos)
Isso indica que o modelo não apenas melhorou na detecção básica, mas também ficou mais preciso em diferentes thresholds de IoU (Intersection over Union).
2. Performance por Classe¶
Bicicletas (maior ganho):
- Precision: 83.3% → 100% (+16.7%)
- Recall: 75% → 96.6% (+21.6%)
- O modelo inicialmente tinha dificuldade com bicicletas devido à estrutura mais complexa (rodas vazadas, diferentes ângulos). Com mais épocas, aprendeu melhor as características distintivas.
Carros (já excelente):
- Manteve Recall perfeito (100%)
- Precision leve queda: 92.2% → 95.2%
- Carros são mais fáceis de detectar: formas sólidas e padronizadas
3. Convergência do Modelo¶
Analisando os gráficos de loss:
- Box Loss: Redução consistente (melhor localização das bounding boxes)
- Obj Loss: Diminuição contínua (melhor detecção de objetos)
- Cls Loss: Próximo de zero (excelente classificação entre as 2 classes)
Não há sinais evidentes de overfitting, pois as curvas de validação acompanham as de treinamento sem divergência.
4. Trade-off: Tempo vs Performance¶
- Tempo: 40 épocas levaram 2.7x mais tempo que 15 épocas
- Ganho: +7.5% em mAP@0.5, +21.5% em mAP@0.5:0.95
- Conclusão: O trade-off é favorável para aplicações que necessitam alta precisão. Para aplicações em tempo real com recursos limitados, 15 épocas já oferece performance aceitável (92% mAP@0.5).
5. Pontos Fortes do Sistema¶
✅ Dataset bem balanceado (4 imagens de cada classe no validation) ✅ Rotulação precisa (corrigida após identificação do erro inicial) ✅ Convergência saudável sem overfitting ✅ Performance final excelente (99.5% mAP@0.5) ✅ Generalizável para cenários similares
6. Limitações Identificadas¶
⚠️ Dataset pequeno: 64 imagens de treino é limitado ⚠️ Diversidade: Imagens parecem ter contextos similares (urbano) ⚠️ Oclusão: Não testado com objetos parcialmente escondidos ⚠️ Condições: Falta variedade de iluminação, clima, ângulos extremos
7. Melhorias Futuras¶
- Data Augmentation: Rotação, flip, mudança de brilho/contraste
- Mais dados: Expandir para 200+ imagens por classe
- Cenários diversos: Noite, chuva, oclusão parcial
- Teste em vídeo: Avaliar performance em tempo real
- Outras classes: Adicionar motos, caminhões, pedestres
🏆 Conclusão Final:
O modelo YOLOv5 customizado demonstrou excelente capacidade de aprendizado, evoluindo de 92% para 99.5% de mAP@0.5 com o aumento de épocas. A arquitetura se mostrou adequada para o problema de detecção de bicicletas vs carros, com resultados que validam sua aplicabilidade em cenários reais de monitoramento e segurança patrimonial conforme proposto pela FarmTech Solutions.
🧪 7. TESTE COM IMAGENS DO CONJUNTO DE TESTE¶
Agora vamos testar o modelo treinado (40 épocas) com as imagens do conjunto de teste.
In [26]:
# Realizar detecção nas imagens de teste usando o modelo de 60 épocas
print("🔍 Executando detecções nas imagens de teste...")
# Verificar modelo
%cd /content/yolov5
!python detect.py \
--weights runs/train/treino_40_epocas/weights/best.pt \
--img 640 \
--conf 0.25 \
--source /content/drive/MyDrive/FIAP/FASE_6/dataset/test/images \
--name teste_final \
--save-txt \
--save-conf
print("✅ Detecções concluídas!")
🔍 Executando detecções nas imagens de teste... /content/yolov5 detect: weights=['runs/train/treino_40_epocas/weights/best.pt'], source=/content/drive/MyDrive/FIAP/FASE_6/dataset/test/images, data=data/coco128.yaml, imgsz=[640, 640], conf_thres=0.25, iou_thres=0.45, max_det=1000, device=, view_img=False, save_txt=True, save_format=0, save_csv=False, save_conf=True, save_crop=False, nosave=False, classes=None, agnostic_nms=False, augment=False, visualize=False, update=False, project=runs/detect, name=teste_final, exist_ok=False, line_thickness=3, hide_labels=False, hide_conf=False, half=False, dnn=False, vid_stride=1 YOLOv5 🚀 v7.0-441-g15c0127a Python-3.12.12 torch-2.8.0+cu126 CPU Fusing layers... Model summary: 157 layers, 7015519 parameters, 0 gradients, 15.8 GFLOPs image 1/8 /content/drive/MyDrive/FIAP/FASE_6/dataset/test/images/bike_037.bmp: 480x640 1 bicicleta, 365.4ms image 2/8 /content/drive/MyDrive/FIAP/FASE_6/dataset/test/images/bike_038.bmp: 480x640 1 bicicleta, 357.3ms image 3/8 /content/drive/MyDrive/FIAP/FASE_6/dataset/test/images/bike_039.bmp: 480x640 1 bicicleta, 323.6ms image 4/8 /content/drive/MyDrive/FIAP/FASE_6/dataset/test/images/bike_040.bmp: 480x640 1 bicicleta, 490.8ms image 5/8 /content/drive/MyDrive/FIAP/FASE_6/dataset/test/images/carsgraz_037.bmp: 480x640 1 carro, 481.7ms image 6/8 /content/drive/MyDrive/FIAP/FASE_6/dataset/test/images/carsgraz_038.bmp: 480x640 1 carro, 326.5ms image 7/8 /content/drive/MyDrive/FIAP/FASE_6/dataset/test/images/carsgraz_039.bmp: 480x640 (no detections), 319.8ms image 8/8 /content/drive/MyDrive/FIAP/FASE_6/dataset/test/images/carsgraz_040.bmp: 480x640 1 carro, 313.6ms Speed: 1.7ms pre-process, 372.3ms inference, 1.2ms NMS per image at shape (1, 3, 640, 640) Results saved to runs/detect/teste_final6 7 labels saved to runs/detect/teste_final6/labels ✅ Detecções concluídas!
Visualizar resultados das detecções¶
In [32]:
import glob
from PIL import Image
import matplotlib.pyplot as plt
test_results_path = 'teste_final6' # aqui tem que ser trocado pelo diretório correto, a depender das tentativas de executação do teste anterior.
imgs = glob.glob(f'runs/detect/${test_results_path}/*.bmp')
for img_path in imgs:
img = Image.open(img_path)
plt.figure(figsize=(10, 6))
plt.imshow(img)
plt.axis('off')
plt.title(img_path.split('/')[-1])
plt.show()
📊 Análise dos Resultados de Teste¶
Performance no conjunto de teste (8 imagens):
Taxa de Detecção Geral: 7/8 objetos detectados (87.5%)
Por Classe:
Bicicletas: 4/4 detectadas (100%) ✅
- Todas as bicicletas foram corretamente identificadas
- Confiança média alta (~0.7-0.8)
Carros: 3/4 detectados (75%) ⚠️
- 1 falha: carsgraz_039.bmp não foi detectado
- Possível causa: ângulo, oclusão ou baixo contraste
Observações:
- O modelo teve excelente performance em bicicletas (100%)
- Carros tiveram 1 falso negativo (25% erro)
- Não houve falsos positivos
- Confiança das detecções: alta (>0.6)
Casos de Sucesso:
- Bicicletas detectadas corretamente mesmo com diferentes ângulos
- Carros identificados em cenários urbanos
Limitações Identificadas:
- 1 carro não detectado (investigar características da imagem)
- Dataset de teste pequeno (apenas 8 imagens)
Conclusão: O modelo demonstrou boa generalização com 87.5% de acerto no teste, validando os resultados de treinamento (mAP@0.5 de 99.5%).